A new fuzzy time series forecasting model combined with ant colony optimization and auto-regression

This paper presents a new fuzzy time series model combined with ant colony optimization (ACO) and auto-regression. The ACO is adopted to obtain a suitable partition of the universe of discourse to promote the forecasting performance. Furthermore, the auto-regression method is adopted instead of the traditional high-order method to make better use of historical information, which is proved to be more practical. To calculate coefficients of different orders, autocorrelation is used to calculate the initial values and then the Levenberg-Marquardt (LM) algorithm is employed to optimize these coefficients. Actual trading data of Taiwan capitalization weighted stock index is used as benchmark data. Computational results show that the proposed model outperforms other existing models.